* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download Monetary and Fiscal Policy with Sovereign Default
Survey
Document related concepts
Financialization wikipedia , lookup
Systemic risk wikipedia , lookup
Debt collection wikipedia , lookup
Debt settlement wikipedia , lookup
Interest rate wikipedia , lookup
Monetary policy wikipedia , lookup
Debt bondage wikipedia , lookup
Quantitative easing wikipedia , lookup
Debtors Anonymous wikipedia , lookup
First Report on the Public Credit wikipedia , lookup
Inflation targeting wikipedia , lookup
Financial correlation wikipedia , lookup
Household debt wikipedia , lookup
Transcript
Monetary and Fiscal Policy with Sovereign Default∗ Joost Röttger† January 13, 2016 Abstract How does the option to default on debt payments affect the conduct of public policy? To answer this question, this paper studies optimal monetary and fiscal policy without commitment in a model with nominal debt and endogenous sovereign default. When the government can default on its debt, public policy changes in the short and the long run relative to a setting without default option. The risk of default increases the volatility of interest rates, impeding the government’s ability to smooth tax distortions across states. It also limits public debt accumulation which reduces the government’s incentive to implement high inflation in the long run. The consequences of having the option to default are found to be negligible in terms of welfare. Keywords: Monetary and Fiscal Policy, Public Debt, Sovereign Default, Lack of Commitment JEL Classification: E31, E63, H63 ∗ First version: October 14, 2012. I thank participants at the 2013 North American Summer Meeting of the Econometric Society (Los Angeles), the 2014 European Meeting of the Econometric Society (Toulouse), the 2013 European Macroeconomics Workshop (London), the 2013 Royal Economic Society Annual Conference (London), the 19th International Conference on Computing in Economics and Finance (Vancouver) and the 2nd Workshop on Financial Market Imperfections and Macroeconomic Performance (Cologne) for helpful comments. Financial support from the RGS Econ and the German Research Foundation (DFG Priority Programme 1578) is gratefully acknowledged. Previous versions of this paper circulated under the titles "Public Debt, Inflation, and Sovereign Default" and "Discretionary Monetary and Fiscal Policy with Endogenous Sovereign Default". † University of Cologne, Center for Macroeconomic Research, Albertus-Magnus-Platz, 50923 Cologne, Germany; Email: [email protected]. 1 Introduction Suppose that a government faces high nominal debt payments that can only be refinanced at high interest rates. If it is not willing (or able) to raise primary surpluses to pay bond holders, there are essentially two options left: inflation and sovereign default. While default and inflation both can lower the real debt burden, there are several differences between these two policy options which make them imperfect substitutes. For example, a government can collect seigniorage when engineering inflation by issuing currency while a default does not generate additional tax revenues. Another difference is that default directly affects the return on government bonds whereas inflation impacts on the return on all nominal assets. Being a continuous variable, inflation can also be adjusted rather easily while the discrete default choice does not offer the same degree of flexibility. Given the distinct roles of money and government bonds for the private sector, default and inflation may also distort economic activity through different channels. The contribution of this paper is to study the implications of allowing a policy maker not only to use standard instruments of monetary and fiscal policy but also to choose outright sovereign default. In particular, it extends previous studies on optimal monetary and fiscal policy with nominal debt that focus on the case of lack of commitment but still assume that the policy maker is always committed to service debt (see e.g. Diaz-Gimenez et al., 2008, Martin, 2009, Niemann et al., 2013a). In the model studied here, a benevolent government jointly chooses monetary and fiscal policy under discretion to finance exogenous government spending in a representative agent cash-credit economy that is subject to productivity shocks.1 More specifically, it sets a labor income tax rate, chooses the money growth rate, issues nominal non-state contingent bonds and decides on whether to repay its outstanding debt or not. The default decision is modeled as a binary choice (see Eaton and Gersovitz, 1981). Following the quantitative sovereign default literature,2 a default is costly because it leads to a deadweight loss of resources that takes the form of a reduction in aggregate productivity and exclusion from financial markets for a random number of periods. As is common in the literature on optimal monetary and fiscal policy, I consider a closed economy. This paper thus contributes to the study of domestic debt default which, despite being a historically recurring phenomenon with severe economic consequences, has not received a lot of attention in the 1 I assume that there is only one policy maker, referred to as the government, who is in charge of both, monetary and fiscal policy. Niemann (2011), Niemann et al. (2013b) and Martin (2015) study time-consistent public policy without sovereign default in models where a central bank and a fiscal authority interact. Röttger (2015) considers a model with independent monetary and fiscal authorities that allows for sovereign default and political frictions. 2 See e.g. Hamann (2004), Aguiar and Gopinath (2006) and Arellano (2008). 1 sovereign default literature (see Reinhart and Rogoff, 2011). In a closed economy, a default does not redistribute resources from foreign lenders to domestic citizens. The government may still choose not to repay its debt to relax its budget constraint and reduce distortionary taxes. The model is calibrated to the Mexican economy which has experienced periods of high inflation and sovereign risk in the recent past. In addition, domestic nominal debt matters for the Mexican government.3 I study the Markov-perfect equilibrium of the public policy problem (see Klein et al., 2008). The government’s decisions hence only depend on the payoff-relevant state of the economy which consists of aggregate productivity, the beginning-of-period public debt position and whether the government is in financial autarky or not. Since the government optimizes sequentially, it cannot commit to future policies and does not internalize how its current decisions affect household expectations in previous periods. However, the government is aware that (expected) future policy will depend on its borrowing decision because it will affect the incentive to reduce the real debt burden via default or inflation in the next period. The option to default thus matters for the government’s response to adverse shocks by allowing it to adjust the real debt burden as well as by affecting the cost of borrowing and thus the attractiveness of debt as a shock absorber. Compared to an otherwise identical economy without default option (or equivalently an economy with prohibitively high costs of default) the availability of sovereign default results in lower average inflation. Since inflation does not reduce the real debt burden when a default takes place, it is lower when default is chosen instead of repayment. However, this direct effect of default on inflation is negligible at a plausible default frequency. The key mechanism that leads inflation to be lower when the default option is available is an indirect one. The attractiveness and hence the probability of default increases with public debt and decreases with aggregate productivity. With default risk, bond prices become more debt elastic in recessions than with only inflation risk and the marginal revenue from debt issuance decreases faster.4 Consequently, the government borrows less which reduces its incentive to implement high inflation rates. Since lower average debt is associated with less inflation, less money is issued and seigniorage revenues decline. The government then has to adjust the primary surplus, leading to a higher labor tax rate in the long run. In the short run, the increased sensitivity of bond prices to productivity shocks and bond issuance that is induced by sovereign risk impedes the government’s ability to smooth tax distortions across states. Relative to an economy without default 3 While the share of nominal domestic debt held by foreign residents has increased in recent years, its average value was only 13.16% for the time period from 2000 to 2014. These numbers are based on publicly available bond data provided by the Mexican central bank. 4 Even without sovereign risk, higher debt issuance leads to higher interest rates by increasing expected inflation. 2 option, tax and inflation rates are thus more volatile, amplifying the impact of productivity shocks on the economy. From a welfare perspective, it is not obvious whether it is desirable to endow the government with the option to default when it cannot commit to future actions.5 As discussed above, the risk of default affects public policy in the short and the long run. With productivity shocks, the government would like to smooth tax distortions by running a budget deficit (surplus) during bad (good) times, following the logic of Barro (1979). Default risk makes debt issuance more expensive in recessions which leads to welfare losses due to more volatile public policy. The long-run implications of sovereign default might however outweigh these costs. As in Martin (2009) and Diaz-Gimenez et al. (2008), the government chooses positive average debt positions because of its lack of commitment and a monetary friction. By increasing the cost of borrowing in recessions, risk of default renders public debt accumulation less attractive, thus avoiding high debt levels and the implementation of high average inflation. A welfare exercise reveals that the counterfactual elimination of sovereign default leads to a negligible welfare loss. For the case of Mexico, lack of commitment to debt service hence is not particularly important from a welfare perspective, Related Literature This paper builds on the literature on optimal Markov-perfect monetary and fiscal policy with nominal government debt. Martin (2009, 2011, 2013) extensively studies the short- and long-run properties of public debt and inflation when the government lacks commitment. Whereas Martin (2009) studies public policy for different environments with a cash-in-advance constraint, the other two papers both use a money search setting. Diaz-Gimenez et al. (2008) show how public policy and welfare depend on whether debt is indexed to inflation or not. Among other things, they find that without commitment welfare can be lower when debt is indexed. In a model with nominal rigidities, Niemann et al. (2013a) show that the presence of lack of commitment and nominal government debt leads to persistent inflation. Despite highlighting the role of lack of commitment for public policy, all of these studies maintain the assumption that there is no commitment problem related to debt repayment and thus abstract from sovereign default. This work is also related to recent papers that study domestic debt default. In a model with incomplete markets and idiosyncratic income risk, D’Erasmo and Mendoza (2013) show that a sovereign default can occur in equilibrium 5 The same is true in the context of consumer default where there exists a trade-off when introducing the option to file bankruptcy. On the one hand, indebted consumers receive the ability to make debt payments state contingent. On the other hand, this flexibility comes at the cost of higher borrowing costs that compensate lenders for the increased risk of default (see e.g. Livshits et al., 2007). 3 as an optimal distributive policy. Pouzo and Presno (2014) extend the incomplete markets model of Aiyagari et al. (2002) by considering a policy maker who cannot commit to debt payments. SosaPadilla (2013) studies Markov-perfect fiscal policy in a model where a sovereign default triggers a banking crisis. All of these papers feature real economies and hence do not discuss monetary policy. This paper also relates to the quantitative sovereign default literature that studies how risk of default affects business cycles in emerging economies.6 With this literature, it shares the assumption of the government’s lack of commitment and the way sovereign default is modeled. Within this literature, the studies that are closest to this paper are Cuadra et al. (2010), Hur et al. (2014) and Du and Schreger (2015). Cuadra et al. (2010) study a production economy with endogenous fiscal policy but abstract from monetary policy and - as is common in the sovereign default literature - look at a small open economy that trades real bonds with foreign investors. Hur et al. (2014) consider an endowment economy with nominal debt, exogenous shocks to inflation and risk-averse investors. They find that the cyclicality of inflation matters for public debt dynamics by affecting risk premia and thus the cost of borrowing. Du and Schreger (2015) study a model of a small open economy where the government borrows in local currency from foreign investors and can reduce its real debt burden by using inflation. Since domestic entrepreneurs have liabilities denominated in foreign currency but earn revenues in local currency, inflation hurts firm balance sheets by depreciating the local currency.7 In independent and contemporaneous work, Sunder-Plassmann (2014) also studies time-consistent public policy for a version of Martin (2009)’s cash-credit economy that allows for a default decision as in Arellano (2008). However, there are several differences between our studies that make them complementary. First, as in Diaz-Gimenez et al. (2008), the focus of her paper is on comparing the long-run properties of a model economy with nominal government debt and an otherwise identical model economy with real government debt. By contrast, I focus on how the ability to default changes the conduct of monetary and fiscal policy in the short and the long run relative to an economy without default, using a model that can replicate short- and long-run properties of the Mexican economy. To model business cycle fluctuations, I study a model with productivity shocks, whereas her model considers government expenditure shocks. Another difference between our two studies is that her model assumes an exogenous and constant labor tax rate, whereas I allow the government to choose the tax rate. 6A recent summary of this literature can be found in Aguiar and Amador (2014). et al. (2015) also develop a quantitative sovereign default model where the government can depreciate the local currency but consider external debt that is denominated in foreign currency. 7 Na 4 Recently, Aguiar et al. (2013) have also developed a model to jointly study inflation and sovereign default when a government cannot commit to future policy. However, their analysis differs from mine in several ways. First, their model features an endowment economy that is not subject to fundamental shocks and borrows from abroad. Second, the authors assume that the government experiences an ad-hoc utility cost of inflation. Third, in the spirit of Cole and Kehoe (2000), they exclusively focus on self-fulfilling debt crises. The rest of the paper is organized as follows. Section 2 presents the model that is analyzed in Section 3. The welfare implications of sovereign default are discussed in Section 4. Section 5 concludes. 2 Model The model extends the cash-credit economy studied in Martin (2009) by allowing for productivity shocks and sovereign default.8 Time is discrete, starts in t = 0 and goes on forever. The economy is populated by a unit mass continuum of homogeneous infinitely-lived households and a benevolent government. Taking government policies and prices as given, the households optimize in a competitive fashion. They supply labor nt to produce the marketable good yt , using a linear technology to be specified below. In addition, they choose consumption of a cash good c1t and a credit good c2t , and rf decide on money (m̃t+1 ), nominal government bond (b̃t+1 ) and nominal risk-free private bond (b̃t+1 ) holdings. The unit price of a government (private) bond is denoted as qt (qtr f ). The risk-free bonds are only traded by households. While all assets are nominal and thus subject to inflation risk, only government bonds are subject to default risk. A role for money is introduced by tying consumption of c1t to beginning-of-period money holdings via a cash-in-advance constraint (see Lucas and Stokey, 1983, Svensson, 1985) m̃t ≥ p̃t c1t , with p̃t denoting the price of consumption in terms of m̃t . To finance constant government spending g and outstanding nominal debt payments B̃t , the government chooses from a set of policies that includes the money growth rate µt , a linear labor income tax rate τt , the binary default decision dt ∈ {0, 1}, and issuance of nominal non-state contingent oneperiod bonds B̃t+1 . A default occurs when dt = 1 is chosen, while the government fully repays its 8 The focus on productivity shocks allows me to study how the possibility of sovereign default affects the business cycle properties of a monetary economy. 5 obligations for dt = 0. In the default case, the government is excluded from financial markets for a random number of periods (see Aguiar and Gopinath, 2006, or Arellano, 2008), in which it can thus neither borrow from nor lend to households, i.e. B̃t+1 = 0. The government’s credit status is given by the indicator variable ht ∈ {0, 1}. If ht = 0, the government has access to the bond market, whereas it is in financial autarky for ht = 1. If the government defaults (dt = 1), its credit status immediately switches to ht = 1. Conditional on being in autarky in a given period t, the government regains access to the bond market in the subsequent period (ht+1 = 0) with probability θ and remains in autarky (ht+1 = 1) with probability 1 − θ . Conditional on not being in autarky, the government will have access to the bond market until it chooses to default. 2.1 Private Sector Households have preferences given by " E0 # ∞ t ∑ β u(c1t , c2t , nt ) , t=0 with discount factor β ∈ (0, 1) and period utility function u : R3+ → R. The utility function is additively separable in all its arguments and satisfies u1 , u2 , −un > 0 and u11 , u22 , unn < 0 with ux (uxx ) denoting the first (second) derivative of u(·) with respect to x ∈ {c1 , c2 , n}. Households have initial assets ∞ and government policies {d , µ , τ , B̃ ∞ (b0 , br0 f , m0 ) and take as given prices { p̃t , qt , qtr f }t=0 t t t t+1 }t=0 . The aggregate money stock evolves according to M̃t+1 = (1 + µt )M̃t . Households also take as given ∞ ∞ the government’s credit status {ht }t=0 . The labor productivity {at }t=0 of the households is subject to random shocks and follows a stationary first-order Markov process with continuous support A ⊆ R+ and transition function fa (at+1 |at ). Households maximize their expected lifetime utility subject to the period budget constraint (1 − τt )ψ(at , ht )nt − c1t − c2t = rf m̃t +b̃t +b̃t p̃t rf rf − rf m̃t +b̃t p̃t the cash-in-advance constraint m̃t ≥ c1t , p̃t 6 m̃t+1 +qt b̃t+1 +qt b̃t+1 , p̃t if ht = 0 if ht = 1 rf rf − m̃t+1 +qt b̃t+1 , p̃t as well as the No-Ponzi game constraint limT →∞ Et h i T −1 r f qrTf b̃rTf+1 + qT b̃T +1 ∏s=0 qt+s ≥ 0. Households use their labor supply nt to produce a marketable good according to the linear technology yt = ψ(at , ht )nt . They take as given their effective labor productivity ψ : R+ × {0, 1} → R+ which depends on random productivity at and the government’s credit status ht (see Cuadra et al., 2010). Effective productivity ψ(·) increases with exogenous productivity (∂ ψ(at , ht )/∂ at ≥ 0) and is negatively affected if the government has a bad credit status (ψ(at , 0) ≥ ψ(at , 1)).9 2.2 Public Sector Conditional on the government’s credit status, the government budget constraint is g − τt ψ(at , ht )nt = M̃t+1 +qt B̃t+1 p̃t − M̃tp̃+t B̃t , M̃t+1 −M̃t , p̃t if ht = 0 if ht = 1 In the default (and autarky) case, the government has to finance public spending g with income tax revenues τt ψ(at , 1)nt and seigniorage τtm ≡ M̃t+1 − M̃t / p̃t . When the government repays its debt, it additionally has to make debt payments but can access the bond market and issue debt. Following the quantitative sovereign default literature (see e.g. Arellano, 2008 and Cuadra et al., 2010), a sovereign default entails two types of costs for the economy. First, the government is excluded from the bond market in the default period and remains in autarky for a random number of periods.10 Second, the economy experiences a direct resource loss governed by ψ(·). As in Cuadra et al. (2010) and Pouzo and Presno (2014), these costs capture in reduced form productivity losses that occur in periods of default (and financial autarky). Despite being arguably ad hoc, such a specification allows me not to take a stand on how exactly a sovereign default is propagated through the economy. While there is evidence for domestic output costs, there is still no consensus on which mechanism is the most relevant one (see Panizza et al., 2009). In addition, two recent papers show that models 9 It is straightforward to modify the model to include a representative firm that is owned by households and produces the homogeneous good yt , using labor supplied by households at a real wage wt . Due to linearity of the production function, the wage rate will equal effective productivity ψ(at , ht ) and profits will be zero, such that the behavior of the economy will not change with such a firm sector. 10 Note that households can still trade risk-free bonds among each other when the government is in financial autarky. However, since they are homogeneous and private bonds are in zero net supply, this is not going to affect the way public policy is conducted. 7 with endogenous default costs that arise due to private credit disruptions (Mendoza and Yue, 2012) or banking crises (Sosa-Padilla, 2013) deliver similar qualitative and quantitative results as those with exogenous default costs. 2.3 Private Sector Equilibrium The first-order conditions for the household problem are given by − un (nt ) u2 (c2t ) = (1 − τt )ψ(at , ht ), p̃t u2 (c2t ) = β Et u1 (c1t+1 ) , p̃t+1 p̃t u2 (c2t )qt = β Et (1 − dt+1 ) u2 (c2t+1 ) , p̃t+1 p̃t u2 (c2t )qtr f = β Et u2 (c2t+1 ) . p̃t+1 (1) (2) (3) (4) In addition, the following complementary slackness conditions need to be satisfied as well: λt = u1 (c1t ) − u2 (c2t ) ≥ 0, m̃t / p̃t − c1t ≥ 0, λt (m̃t / p̃t − c1t ) = 0, with λt denoting the Kuhn-Tucker multiplier on the cash-in-advance constraint.11 Intuitively, the cash-in-advance constraint is binding whenever the marginal utility of cash-good consumption exceeds the marginal utility of credit-good consumption. The inequality u1 (c1t ) − u2 (c2t ) ≥ 0, (5) needs to hold in equilibrium to satisfy λt ≥ 0. Equation (1) characterizes the optimal household labor supply decision which is distorted for non-zero tax rates τt 6= 0. The conditions (2)-(4) are the Euler equations for money and bonds.12 Since nominal government bonds are defaultable, they have to compensate households for the risk of default (see condition (3)). However, all assets need to compensate for expected (gross) inflation p̃t+1 / p̃t . As in Martin (2009), I normalize nominal variables by the beginning-of-period aggregate money stock M̃t , xt ≡ x̃t /M̃t for x ∈ B, b, br f , m, p , which renders the model stationary.13 It implies that the 11 In a household optimum, the household budget constraint and the No-Ponzi game constraint hold with equality. 12 The Euler equation for government bonds (3) only holds for periods in which the government is not in financial autarky. 13 Note that, by construction, the normalized aggregate money stock is constant and equal to one. 8 inflation rate in period t is given as πt ≡ pt (1 + µt−1 ) − 1, pt−1 such that inflation equals money growth in the long run. After normalizing nominal variables, the Euler equations are now given as pt 1 , u2 (c2t ) = β Et u1 (c1t+1 ) pt+1 1 + µt pt 1 u2 (c2t )qt = β Et (1 − dt+1 ) u2 (c2t+1 ) , pt+1 1 + µt 1 pt rf . u2 (c2t )qt = β Et u2 (c2t+1 ) pt+1 1 + µt (6) (7) (8) For the economy, the goods and asset market clearing conditions are as follows: ψ(at , ht )nt = c1t + c2t + g, bt+1 = Bt+1 , rf bt+1 = 0, mt+1 = 1. If real balances are high enough, i.e. 1/pt ≥ u−1 1 (u2 (c2t )) holds (see condition (5)), households equalize marginal utility across cash and credit goods. If not, households are cash constrained and the allocation of consumption is distorted. As in Martin (2009), in a monetary equilibrium, i.e. an equilibrium in which money is valued, c1t = 1/pt , needs to hold. Note that this still allows for an unconstrained consumption allocation if the cash-inadvance constraint is just binding, i.e. when pt is such that λt = 0 and c1t = 1/pt = u−1 1 (u2 (c2t )). 2.4 Public Policy Problem In this section, I formulate the public policy problem. The government is benevolent and sets its policy instruments to maximize the expected life-time utility of the households, anticipating the response of the private sector to its policies. However, it cannot commit itself to a state-contingent (Ramsey) policy plan for all current and future policies but optimizes from period to period instead. 9 To analyze the decision problem of the government, I restrict attention to stationary Markov-perfect equilibria (see Klein et al., 2008). In a Markov-perfect equilibrium, the optimal decisions of the government in any period will be characterized by time-invariant functions that only depend on the minimal payoff-relevant state of the economy in that respective period. In the model, this state consists of the beginning-of-period debt-to-money ratio Bt , labor productivity at and the government’s credit status ht . By requiring the government to only condition its decisions on the current payoffrelevant aggregate state, the Markov-perfect equilibrium concept rules out the possibility that the government is able to keep promises made in the past. This is because at the start of a period, the government does not care about the past and only considers its payoff in current and future periods.14 By construction, the government thus is ensured to act in a time-consistent way. The Markov-perfect policy problem will be formulated recursively. In the remainder, I will thus adopt the notation of dynamic programming. Time indices are hence dropped and a prime is used to denote next period’s variables. Given the aggregate state at the start of a period, the government takes as given the policy function D(B0 , a0 ) that determines next period’s default decision as well as the policy functions X r (B0 , a0 ) and X d (a0 ), with X ∈ {C2 , P}, that determine consumption and the price index in the next period for the case of repayment (r) and default (d).15 Expectations of these variables enter the household optimality conditions (6) and (7) and thus matter for the allocation in the current period.16 Despite lacking the ability to commit to future policies, the government fully recognizes today that it affects (expected) future policies via its choice of B0 , which in turn have an effect on the behavior of the private sector in the current period. In a stationary Markov-perfect equilibrium, the policy functions that govern future decisions then coincide with the policy functions that determine current public policy for all states. As in Klein et al. (2008), one can interpret the formulation of the public policy problem as a Markov-perfect game played between successive governments. Following this interpretation, in each period, a different government is in charge of choosing public policy. Each government then chooses its optimal strategies, taking as given the optimal responses of the government in the next period. In every period, the government anticipates how the private sector responds to its actions as given by the private sector equilibrium conditions.17 Applying the normalization of nominal variables used 14 The focus on Markov-perfect strategies also rules out the possibility of reputational considerations based on complex trigger strategies as in Chari and Kehoe (1990, 1993). 15 Remember that cash-good consumption c is directly linked to the price index p via the cash-in-advance constraint. 1 16 Households do not have a strategic impact on future government policies but form rational expectations about them based on the policy functions listed above. 17 The government thus plays a Stackelberg game against the (passive) private sector in every period. 10 earlier, the government budget constraint can be written as g − τt ψ(at , ht )nt = (1 + µt ) 1+qt Bt+1 − 1+Bt , if pt pt µt pt , if ht = 0 ht = 1 where M̃t+1 = (1 + µt )M̃t is used as well. Using the household optimality conditions (1),(6)-(7), the binding cash-in-advance constraint and the aggregate resource constraint, the government budget constraint can be further rewritten as 0 0 r (B0 , a0 )−1 ) 1−D(B ,a ) h i u (P 0 0 1 P r (B0 ,a0 ) r 0 0 1−D(B ,a ) B0 0 |a u2 (C (B , a )) β Ea0 |a + β E r 0 0 a 2 P (B ,a ) 0 ,a0 ) +u1 (P d (a0 )−1 ) D(B P d (a0 ) (9) +un (n)n + u2 (c2 ) (c2 − B/p) = 0, for the repayment case and as β Ea0 |a θ × 0) u1 (P r (0, a0 )−1 ) 1−D(0,a P r (0,a0 ) 0) +u1 (P d (a0 )−1 ) D(0,a P d (a0 ) + (1 − θ ) × u1 (C1d (a0 )) P d (a0 ) (10) +un (n)n + u2 (c2 )c2 = 0, for the default (and autarky) case. This constraint can be seen as the period implementability constraint for the government.18 However, in addition to this constraint, the government has to respect the following two private sector equilibrium conditions: 0 = ψ(a, h)n − 1/p − c2 − g, (11) 0 ≤ u1 (1/p) − u2 (c2 ). (12) The household budget constraint is satisfied by Walras’ Law, given the government budget constraint, the binding cash-in-advance constraint and the market clearing conditions. Let B ≡ [B, B] be the set of possible aggregate debt values with −∞ < B ≤ 0 and 0 < B < ∞. Conditional on entering a period with access to financial markets, the decision problem of the government is given by the following functional equation: n o r d V(B, a) = max (1 − d)V (B, a) + dV (a) d∈{0,1} 18 The derivation of the implementability constraint can be found in Appendix A.1. 11 (13) with the value of repayment given as V r (B, a) = max0 c2 ,n,p,B ∈B u (1/p, c2 , n) + β Ea0 |a V(B0 , a0 ) s.t. (9), (11), (12), and the value of default as h i V d (a) = max u (1/p, c2 , n) + β Ea0 |a θ V(0, a0 ) + (1 − θ ) V d (a0 ) s.t. (10) − (12). c2 ,n,p The value V(·) is the option value of default. As is standard in the sovereign default literature, the government is assumed to honor its obligations whenever it is indifferent between default and repayment. If the government starts a period in financial autarky, it solves the same problem as in the default case. When in autarky, the government will regain access to financial markets in the subsequent period with probability θ . With probability 1 − θ , it will stay in financial autarky.19 2.5 Equilibrium The Markov-perfect equilibrium is defined as follows: Definition 1. A stationary Markov-perfect equilibrium consists of two sets of functions {D, B r , C2r , N r , P r , V, V r } : B × A → {0, 1} × B × R3+ × R2 and {C2d , N d , P d , V d } : A → R3+ × R, such that for all (B, a) ∈ B × A, D (B, a) = arg max o n (1 − d)V r (B, a) + dV d (a) , d∈{0,1} {X r (B, a)}X ∈{C2 ,N ,P,B} = arg max u (1/p, c2 , n) + β Ea0 |a V(B0 , a0 ) c2 ,n,p,B0 ∈B s.t. (9), (11), (12), and h i {X d (a)}X ∈{C2 ,N ,P} = arg max u (1/p, c2 , n) + β Ea0 |a θ V(0, a0 ) + (1 − θ ) V d (a0 ) c2 ,n,p s.t. (10) − (12), as well as V(B, a) = (1 − D (B, a)) × V r (B, a) + D (B, a) × V d (a), 19 On average, the government thus spends 1/θ periods in autarky after a default. 12 V r (B, a) = u(P r (B, a)−1 , C2r (B, a) , N r (B, a)) + β Ea0 |a V(B r (B, a) , a0 ) , and h i V d (a) = u(P d (a)−1 , C2d (a) , N d (a)) + β Ea0 |a θ V(0, a0 ) + (1 − θ ) V d (a0 ) . This definition highlights the stationarity of the policy problem since the functions that solve the decision problem of the government in a given period coincide with the policy functions that govern the optimal decisions of the government in future periods. 3 Model Analysis In this section, the role of sovereign default for public policy is investigated. Because the model cannot be evaluated analytically due to the discrete default option, numerical methods are applied. Appendix A.3 contains details regarding the numerical computation of the equilibrium. The next section presents the model specification. A discussion of the public policy trade-offs can be found in Section 3.2. Simulation results are presented in Section 3.3. 3.1 Model Specification To explore the model properties by computational means, functional forms and parameters need to be chosen. 3.1.1 Functional Forms Productivity follows a log-normal AR(1)-process, i.i.d. ρ at = at−1 exp (σ εt ) , εt ∼ N(0, 1). The household utility function is specified as u(c1 , c2 , n) = γ1 1 c1−σ −1 c1−σ2 − 1 (1 − n)1−σn − 1 1 + γ2 2 + (1 − γ1 − γ2 ) , 1 − σ1 1 − σ2 1 − σn with γ1 , γ2 , σi > 0, i ∈ {1, 2, n} and γ1 + γ2 < 1.20 20 For σi = 1, i ∈ {1, 2, n}, household utility is logarithmic for the respective variable. 13 The resource costs of default are specified as in Cuadra et al. (2010): ψ(a, h) = a − h × max {0, a − ã} . If the government is in default/autarky, effective productivity equals ã when a exceeds ã while there are no costs of default when productivity a is below the threshold ã. This default cost specification implies that a default is more costly in booms than in recessions. In the quantitative sovereign default literature, it is well known that this feature is crucial for default to mostly take place in bad states and hence for countercyclical sovereign risk to emerge (see e.g. Aguiar and Amador, 2014). This property is consistent with empirical evidence (see Tomz and Wright, 2007) and also present in models with endogenous costs of default (see Mendoza and Yue, 2012, Sosa-Padilla, 2013).21 3.1.2 Parameters A model period corresponds to one quarter. The selected model parameters are listed in Table 1. They are either set to standard values or chosen to replicate certain short- or long-run properties of the Mexican economy.22 The productivity parameter ρ is set to 0.9 while σ is chosen to match the standard deviation of HP-filtered Mexican log real GDP. As is common in business cycle models, a discount factor of β = 0.99 is selected, implying an annual real risk-free rate of 4%. Based on World Bank data for 1980-2008, g is set to 0.0379 to match an average ratio of public spending to GDP of around 11%.23 The credit-good preference parameter σ2 is normalized to 1. Targeting a cash-credit good ratio and an average working time of one third each, γ1 is set to 0.003 and γ2 to 0.337. For the inverse of the elasticity of leisure σl , a rather standard value of 2 is selected. The probability of reentry θ is set to 0.2, implying that financial autarky lasts for 5 quarters on average. This parameter value is in line with values considered in the quantitative sovereign default literature which range from 0.0385 (Chatterjee and Eyigungor, 2012) to 0.282 (Arellano, 2008). The size of σ1 is crucial for long-run debt and inflation as will be explained in the next section. To match the Mexican average annual inflation rate of 29.69%, σ1 is set to 2.43.24 The incentive 21 Allowing for default costs that enter the the aggregate resource constraint (or the government budget constraint) in a lump-sum way does not change the results of this paper as long as these losses are also relatively higher in good than in bad states, preserving countercyclical default incentives. 22 The time series for real GDP and the GDP deflator are taken from Cuadra et al. (2010) and cover the time period from 1980:I to 2007:I. They are seasonally adjusted via EViews’ multiplicative X-12 routine. 23 More specifically, I use annual data from the World Bank’s World Development Indicators for general final government consumption as a share of GDP. 24 The inflation rate is calculated based on the quarterly GDP deflator time series for Mexico provided by Cuadra et al. (2010). Using alternative measures such as the CPI also yields average inflation rates of around 30%. 14 Parameter β g γ1 γ2 σ1 σ2 σn ã θ σ ρ Description Discount factor Government spending Cash-good weight Credit-good weight Cash-good curvature Credit-good curvature Leisure curvature Default cost parameter Probability of reentry Std. dev. productivity shock Persistence of productivity Value 0.9900 0.0379 0.0030 0.3370 2.4300 1.0000 2.0000 0.9900 0.2000 0.0169 0.9000 Table 1: Parameter values to default critically depends on ã. For Mexico, Reinhart (2010) documents that there have been domestic defaults in 1982 and between 1929 and 1938. Based on this observation, I set the default cost parameter to match an annual default frequency of 2%. The model is also solved and simulated with prohibitively high productivity costs of default which rule out equilibrium default. This benchmark economy yields the same results as a model without default option and will be referred to as "nodefault economy". The model with default option will be referred to as "baseline economy". 3.2 Public Policy Decisions The optimal policies for the economy with default can be seen in Figure 1. Debt is normalized by nominal output Y ≡ py, evaluated at the unconditional mean of productivity. The default decision is visualized using the default threshold â(B) which is the lowest productivity level that leads to repayment for given debt B: V r (B, â(B)) = V d (â(B)). The threshold separates the state space (B, a) into two areas: the default region (a < â(B), i.e. below the line) and the repayment region (a ≥ â(B), i.e. on and above the line). As in the quantitative sovereign default literature, default becomes more attractive with higher debt and lower productivity (see Arellano, 2008). The remaining policies are presented for productivity levels 1.5 standard deviations below (dashed line) and above (solid line) the unconditional mean of productivity. The nominal interest rate is defined as i = 1/q − 1. Since the continuous policy decisions depend on the default decision, the objects displayed in Figure 1 exhibit kinks at states where default is optimally chosen.25 In the default case, the policies also do not change with B anymore. 25 The policy functions displayed in Figure 1 are given as X (B, a) = (1 − D(B, a))X r (B, a) + D(B, a)X d (a), with X ∈ {B, C2 , N , P}. The remaining variables (µ, τ, i) are calculated similarly by using the policy functions for the default and repayment case and the private sector equilibrium conditions (1), (6)-(7). 15 0.25 0.4 0.2 0.2 0.15 0 −0.2 0 0.1 0.1 0.2 0.8 0.3 0.2 B/Y Real balances 0 0 0.3 0.08 0.075 0.07 0.2 0.2 0.1 1/p 0.4 i 0.2 0.1 0.05 0 0.3 B/Y Nominal interest rate 0.6 0 0 τ 0.6 0.3 0.1 Tax rate Money growth rate μ Default threshold 0.98 0.96 Repayment 0.94 0.92 0.9 Default 0.88 0.86 0 0.1 0.2 0.3 B/Y Borrowing 0.4 B’/Y â Figure 1: Default threshold and selected policy functions (baseline economy) 0.1 B/Y 0.2 0.3 alow ahigh 0.065 0 0.1 B/Y 0.2 0.3 B/Y Notes: The policy functions are depicted for productivity values 1.5 standard deviations below (alow ) and above (ahigh ) the unconditional mean of productivity. The money growth rate and the nominal interest rate are displayed in annual terms. The optimal labor and inflation tax distortions reflect the government’s financing needs. By relaxing the government’s budget, a sovereign default allows to reduce labor taxation and increase real balances relative to full debt repayment. The income tax rate and the price index p both increase with B. An inflationary monetary policy becomes particularly more attractive with higher debt because it lowers the real debt burden. This implies that default and inflation are substitutes since inflation as "partial default" becomes useless for d = 1. However, they are only imperfect substitutes due to the discrete nature of default. The intertemporal policy trade-off can be illustrated via the generalized Euler equation Z ∞ β â(B0 ) u2 (c02 ) ξ −ξ fa (a0 |a)da0 = ξ p0 0 ∂ Rb 0 ∂ Rm B+ ∂ B0 ∂ B0 , (14) where ξ denotes the multiplier on the implementability constraint (9), Rb = ((1 + µ) qu2 ) /p average revenues from bond issuance and Rm = ((1 + µ) u2 ) /p (gross) revenues from money creation.26 Note that these revenues are weighted by the marginal utility of credit-good consumption u2 (c2 ) and 26 The derivation of the Euler equation assumes differentiability of V r , Rb and Rm with respect to debt (see Appendix A.2 for details). As is common in the sovereign default literature (see e.g. Cuadra et al., 2010, or Hatchondo et al., 2015), the generalized Euler equation is only presented here to highlight the intertemporal public policy trade-off in an intuitive way. The numerical algorithm that is used to solve the model is not based on this Euler equation and does not require differentiability to hold (see Appendix A.3 for details). 16 expressed in real terms. Using the household first-order conditions for government bond and money holdings, these revenues can be expressed as the following functions of productivity a and borrowing B0 : u2 (C2r (B0 , a0 )) fa (a0 |a)da0 , r (B0 , a0 ) 0 P â(B ) (Z 0 ) Z ∞ r (B0 , a0 )−1 â(B ) u P d (a0 )−1 u P 1 1 Rm (B0 , a) = β × fa (a0 |a)da0 + fa (a0 |a)da0 . P d (a0 ) P r (B0 , a0 ) 0 â(B0 ) b 0 Z ∞ R (B , a) = β Households dislike volatile consumption and leisure. In the presence of productivity shocks, the government can issue debt to accommodate these preferences and smooth tax distortions as measured by ξ across states (see the LHS of (14)). Its ability to do so is constrained by financial market incompleteness and lack of commitment. Since only nominal non-state contingent bonds are available, the government has an incentive to make real debt payments state contingent via inflation or default. However, because it cannot commit to a state-contingent repayment plan for the next period, public financing conditions will depend on the chosen debt position B0 since it affects the risk of inflation and default. The derivatives on the RHS of (14) reflect this channel. The optimal debt policy then trades off the government’s tax smoothing motive against the time-inconsistency problem, i.e. how current debt issuance affects revenues Rb (B0 , a) × B0 and Rm (B0 , a) by changing household expectations of inflation and default, which matter for the households’ bond and money demand. The impact of debt issuance on public revenues is visualized in Figure 2. It depicts the functions Rb and Rm for the no-default economy (first row) and the baseline economy with equilibrium default (second row). The no-default case has previously been studied by Martin (2009, 2011, 2013) and Diaz-Gimenez et al. (2008) and is hence well understood. It is useful to first revisit this case to understand how sovereign risk affects the debt policy.27 The shape of Rb reflects the relation between inflation and beginning-of-period debt (see panel (a)). Due to its lack of commitment, the government optimizes from period to period and therefore does not internalize how its current policy choices affect outcomes in previous periods. More specifically, it does not internalize that its current actions have an impact on households’ demand for money and bonds in the last period. Failing to recognize this impact, the government decides to erode the real value of beginning-of-period debt via inflation to relax its budget. Since the temptation to use inflation in this way increases with B, expected inflation becomes an increasing function 27 In this case, the default threshold is given as â(B) = 0, i.e. default is prohibitively costly. 17 0.14 Rm R b 0.14 0.12 0.1 0.1 0 ln(a) −0.1 0 0.12 0.1 0.1 0.4 0 0.2 B’/Y ln(a) −0.1 0 (b) Rm (B0 , a) (no default) (a) Rb (B0 , a) (no default) 0.15 0.14 Rm Rb 0.4 0.2 B’/Y 0.075 0 0.1 0 ln(a) −0.1 0 0.4 0.2 B’/Y 0.12 0.1 0.1 0 ln(a) −0.1 0 0.4 0.2 B’/Y (d) Rm (B0 , a) (baseline) (c) Rb (B0 , a) (baseline) Figure 2: Average bond revenues Rb (B0 , a) and money revenues Rm (B0 , a) for the no-default economy and the baseline economy of end-of-period debt and the price of nominal government bonds responds to B0 in a negative way, causing borrowing to become more expensive when more debt is issued (∂ Rb /∂ B0 < 0).28 The shape of Rm reflects the way public debt, inflation and household money demand are related (see panel (b)). Given that real balances 1/p are a decreasing function of B, the real value of money holdings is expected to decrease with borrowing B0 , reducing household money demand today. However, lower real balances 1/p also increase the marginal utility of cash-good consumption u1 (1/p), such that the demand for money increases with B0 since households expect to be more cash-constrained in the subsequent period. Whether higher borrowing B0 increases net household money demand depends on the relative size of the two effects. With last section’s specification of the household utility function, 28 When credit consumption c2 decreases with debt, there is a potential additional effect that might improve borrowing conditions since u2 (c2 ) would increase with B in this case (see Krusell et al., 2005). However, I find credit consumption to be (mildly) increasing with debt for the chosen calibration. 18 Figure 3: Selected policy functions (no-default economy) 0.4 0.01 0.2 0 0.085 0.2 0.3 −0.2 0 0.1 0.2 B/Y Nominal interest rate 0.4 0.8 0.3 0.6 0.2 0.4 0.1 0.2 0.3 0 0 0.1 0.2 B/Y Real balances 0.3 0.08 0.075 0.07 0.2 0.1 0.08 0 0.3 1/p 0.1 i B’/Y 0.09 0 B/Y Borrowing 0 0 0.095 τ 0.02 −0.01 0 Tax rate Money growth rate 0.6 μ τm/y Seigniorage 0.03 0.1 B/Y 0.2 B/Y 0.3 0.065 0 alow ahigh 0.1 0.2 0.3 B/Y Notes: The policy functions are depicted for productivity values 1.5 standard deviations below (alow ) and above (ahigh ) the unconditional mean of productivity. The money growth rate and the nominal interest rate are displayed in annual terms. u1 (1/p)/p = γ1 pσ1 −1 holds and the sign (and size) of the net effect is governed by the parameter σ1 .29 For σ1 > 1, an increase in B0 raises household money demand and hence money revenues (∂ Rm /∂ B0 > 0). Since a household’s valuation of money increases with B0 , the government can simply issue more currency to implement a particular price index and thus obtain more revenues from money issuance. The long-run debt position is determined by the two effects described above. A positive sign for ∂ Rm /∂ B0 is crucial for non-zero long-run debt.30 For σ1 = 1, money revenues do not respond to borrowing (∂ Rm /∂ B0 = 0) which eliminates the incentive to borrow in the long run. With σ1 ∈ (0, 1), the government even has an incentive to accumulate assets (B0 < 0) due to ∂ Rm /∂ B0 < 0.31 These two cases are not further discussed here because they make default a (rather) redundant policy option. When there are productivity shocks, a positive response of money revenues to borrowing also matters for the government’s ability to smooth tax distortions across states since, without this effect, only 29 The parameter γ1 also affects the magnitude of the net effect but not its sign. 30 Looking at Markov-perfect public policy in a real economy setting with endogenous government spending and without default, Debortoli and Nunes (2013) show - for analytical and quantitative examples - that long-run debt only deviates from zero for a small range of parameter values. Similar results are found by Kursell et al. (2005) for a related model with exogenous government spending. 31 For more details see Proposition 5 in Martin (2009) who shows these properties for a deterministic model without default option, assuming that u2 is constant. 19 the negative bond price effect would be operative and make it more expensive to issue debt in low productivity states. When the government can default on its debt, sovereign risk changes the impact of borrowing on public revenues. Panel (c) displays Rb for the baseline economy with sovereign default. Because a default is more likely for higher amounts of debt, the bond price would respond to B0 in a negative way even in the absence of inflation risk. As in Arellano (2008), the debt elasticity of interest rates is higher in bad (low productivity) than in good (high productivity) states, reflecting the default incentives of the government. Panel (d) shows how money revenues respond to B0 when there is sovereign risk. The non-monotonic shape results from the optimal mix of default and inflation. Since default and inflation are substitutes, higher borrowing can lower expected inflation by increasing the probability of default. While making money as a store of value more valuable, this interaction also lowers the expected marginal utility of cash-good consumption and hence Rm . Due to its adverse effect on bond and money revenues, sovereign risk ultimately makes debt issuance less attractive. The consequences for the long-run debt position can be illustrated via Figure 1 and Figure 3. By looking at the intersection between the (dotted) 45-degree line and the borrowing policies, one can already see without having simulated the model that average debt is going to be lower in the baseline model with default. The quantitative dimension of the model properties discussed so far is explored in the next section. 3.3 Simulation Results Table 2 presents the averages of statistics calculated for 2500 simulated economies with 2000 periods each. The time series are filtered using the Hodrick-Prescott filter with a smoothing parameter of 1600. Output is in logs. All simulations are initialized with S0 = (0, E[a]) and the first 500 observations of each sample are discarded. Average debt and inflation are lower with default option. More specifically, the average inflation rate in the no-default economy is more than twice as large as in the baseline economy with default. The possibility of default reduces average inflation through a direct and an indirect effect. When the government chooses to default, there is no incentive to use inflation to reduce the real debt burden anymore. As a result, inflation is lower in default (and autarky) periods than in periods of repayment. The role of this direct effect is however limited by the frequency of default and does not contribute much to the average inflation rate.32 The indirect effect of default on inflation is related to how the 32 The average inflation rate for periods of repayment only is 31.08% and thus only slightly higher than the overall average inflation rate of 30.11% which includes periods of default and autarky as well. 20 Mean Default probability (annual) Debt-to-GDP Tax rate Seigniorage-to-GDP Inflation rate (annual) Nominal interest rate (annual) Standard deviation Output Tax rate Inflation rate (annual) Nominal interest rate (annual) Correlation with output Tax rate Inflation rate (annual) Nominal interest rate (annual) Baseline No default 0.0206 0.1866 0.1043 0.0155 0.3011 0.3958 0 0.3340 0.0912 0.0296 0.6503 0.7175 0.0236 0.0124 0.1077 0.0782 0.0192 0.0012 0.0544 0.0252 -0.8597 -0.2990 -0.4725 -0.9991 -0.5599 -0.9093 Table 2: Selected model statistics risk of default affects the government’s borrowing behavior. Default risk raises the cost of rolling over even low amounts of debt in recessions. This mechanism restricts the build up of large public debt positions which would make higher inflation more attractive. Less debt also implies that the tax base of the income tax increases relative to that of inflation. Hence, the benefit of raising inflation is lower, leading to a higher average labor tax rate in the baseline economy.33 While the accumulation of debt crucially depends on the government’s ability to collect seigniorage (see the discussion in the previous section), the average seigniorage-to-GDP ratio is moderate and of plausible size.34 When the government has the option to default, borrowing is more expensive in recessions due to the increased risk of default. The average nominal interest rate however is higher in the no-default economy since it experiences more inflation on average. The short-run implications of sovereign risk for public policy can be illustrated via Figure 4. For the baseline and the no-default economy, it displays impulse responses of selected model variables to a negative one-time productivity shock. The variables are expressed as absolute deviations from their values at the stationary state to which the economies converge when productivity is kept fixed at its long-run mean.35 Since productivity is persistent (ρ > 0), the negative shock immediately 33 For Mexico, Ilzetzki (2011) calculates an average marginal income tax rate of 12.1% which is close to the average tax rate in the baseline model (10.43%). 34 Using the same definition of seigniorage as in the model, Aisen and Veiga (2008) calculate that average seigniorage is 2.2% of GDP for Mexico. 35 Of course, the two economies do not exhibit the same stationary state. The variables in such a stationary state are close to the average values listed in Table 2. 21 Figure 4: Impulse responses of selected model variables to a negative one-time productivity shock Productivity Borrowing 0 Nominal interest rate 0.5 4 0 −0.5 2 −0.5 0 −1 −1 0 20 Period Default probability 40 2 −1.5 0 20 Period Inflation rate 40 10 1.5 0.8 0.6 5 1 20 Period Tax rate 40 Baseline No default 0.4 0 0.5 0 0 −2 0 20 Period 40 −5 0 0.2 20 Period 40 0 0 20 Period 40 Notes: All variables are expressed as absolute deviations (multiplied by 100) from their values at the stationary state to which the respective economy converges when productivity is kept fixed at its unconditional mean. raises the risk of default as the incentive to default is more likely to be strong in the subsequent period. The high sensitivity of interest rates to changes in debt issuance forces the government to reduce its debt position in order to avoid an even larger decline of the bond price. As a result, the government has to resort to large increases in inflation and taxes to finance debt payments and government spending.36 As productivity reverts back to its mean and debt is reduced even further, expected inflation and sovereign risk both decline, leading the government to take advantage of the improved borrowing conditions and accumulate debt again. In the no-default economy, borrowing conditions do not deteriorate very much in response to the negative productivity shock. This property allows the government to effectively smooth tax distortions across states by issuing debt which avoids large increases in taxes and inflation. Because debt cannot be easily rolled over in the baseline model, the impact of productivity shocks on the economy is more pronounced and output volatility is 23% higher relative to the model without default. Since one of the main contributions of this paper is to offer a joint analysis of inflation and sovereign default, it is interesting to compare the cyclical properties of inflation generated by the 36 This mechanism is related to the one studied by Cuadra et al. (2010) in a model of a small open economy with real government debt. The authors show that countercyclical default risk can rationalize the procyclical consumption taxation observed in emerging economies. 22 Standard deviation Correlation with output Mexico (1980-2007) 0.2423 -0.2734 Baseline 0.1077 -0.2990 No default (recalibrated) 0.0703 -0.4585 Table 3: Cyclical properties of inflation model with and without default to those observed in the data. Table 3 shows the results. In Mexico, inflation is very volatile and countercyclical. The baseline model with default can replicate these findings. To give the model without default a fair chance, it is recalibrated to match the average inflation rate and the volatility of output in Mexico.37 While the recalibrated no-default economy yields countercyclical inflation, the baseline model predicts more volatile and less countercyclical inflation than the no-default model which is closer to what is observed empirically. 4 The Welfare Implications of Sovereign Default This section discusses the welfare implications of sovereign default. With commitment, the option to default will not decrease welfare since the government would otherwise refrain from using it.38 Without commitment, this is not necessarily the case anymore. Section 3 has shown that the default option has implications for public policy in the short and the long run. On the one hand, by increasing the sensitivity of bond prices with respect to debt and productivity, countercyclical risk of default entails short-run costs because the government loses some of its ability to smooth tax distortions across states. On the other hand, default risk might lead to welfare gains due to its impact on long-run debt. The model features a long-run borrowing motive that stems from the presence of two frictions, lack of commitment and a liquidity constraint (see the discussion in Section 3.2). The government acts in a time-consistent way and does not internalize the effect of its current choice of inflation on the borrowing behavior in previous periods. When household money demand and thus the value of money are increasing in the amount of issued debt, the government persistently chooses positive debt positions which then lead to high average inflation. By limiting public debt accumulation via more sensitive interest rates, the default option reduces average inflation and the misallocation of consumption compared to the no-default setting. To evaluate whether the addition of the default option to the set of policy instruments is welfare 37 The changed model parameter values are γ = 0.015, γ = 0.325, σ = 1.73 and σ = 0.02. 1 2 1 38 For a real small open economy with incomplete markets and costly sovereign default, Adam and Grill (2012) show that welfare can be increased when the Ramsey planner can commit to a state-contingent default plan. 23 enhancing, welfare measure ∆ is calculated. It measures the percentage increase in credit-good consumption that households in the no-default economy need to be given in each period to achieve the same expected lifetime utility as in the baseline economy with default: " E0 ∑β " # T t D D u(cD 1t , c2t , nt ) = E0 t=0 # T ∑β t u(cN1t , cN2t (1 + ∆), ntN ) . t=0 The sequences of consumption and labor supply in the economy with ( j = D) and without default T . Expected lifetime utility is calculated for both types option ( j = N) are denoted as {c1tj , c2tj , ntj }t=0 of economies by averaging realized lifetime utility of 2500 samples with simulated time series of effective length T = 1500 each. The calculated welfare measure is ∆ = 0.00006. For the no-default economy, credit-good consumption thus needs to be increased by only 0.006% in each period to equalize household welfare in both types of economies. Since these gains are of negligible size, one can argue that, from a welfare perspective, lack of commitment to repayment is not particularly important for the case of Mexico. 5 Conclusion To understand the implications of the option to default on debt payments for public policy, this paper has studied optimal monetary and fiscal policy without commitment for a cash-credit economy with nominal debt and endogenous government default. While a default allows the government to reduce inflation and distortionary labor taxation by relaxing its budget constraint, the default option mainly induces lower rates of inflation by constraining debt issuance via endogenous default risk premia. This mechanism reduces the average debt position and the government’s incentive to implement high inflation in the long-run. Less debt also implies that the income tax becomes more attractive relative to inflation, resulting in a higher average labor tax rate. Taxes and inflation are more volatile when the default option is available because the government’s ability to smooth tax distortions across states is reduced by the presence of default risk. For the case of Mexico, a counterfactual exercise has demonstrated that the consequences of the option to default for welfare are negligible. 24 References A DAM , K. AND M. G RILL (2012): "Optimal Sovereign Default," Mimeo, University of Mannheim. AGUIAR , M. AND M. A MADOR (2014): "Sovereign Debt," in Handbook of International Economics, Vol. 4, ed. by Helpman, E., K.S. Rogoff and G. Gopinath, North-Holland, Amsterdam, 1243-1277. AGUIAR , M., M. A MADOR , E. FARHI , AND G. G OPINATH (2013): "Crisis and Commitment: Inflation Credibility and the Vulnerability to Sovereign Debt Crises," Mimeo, Princeton University. AGUIAR , M. AND G. G OPINATH (2006): "Defaultable Debt, Interest Rates and the Current Account," Journal of International Economics, 69(1), 64-83. A ISEN , A. AND F.J. V EIGA (2008): "The Political Economy of Seigniorage," Journal of Development Economics, 87(1), 29-50. A IYAGARI , S.R., R. M ARCET, T.J. S ARGENT, AND J. S EPPÄLÄ (2002): "Optimal Taxation without State-Contingent Debt," Journal of Political Economy, 110(6), 1220-1254. A RELLANO , C. (2008): "Default Risk and Income Fluctuations in Emerging Economies," American Economic Review, 98(3), 690-712. A RELLANO , C. AND A. R AMANARAYANAN (2012): "Default and the Maturity Structure in Sovereign Bonds," Journal of Political Economy, 120(2), 187-232. BARRO , R.J. (1979): "On the Determination of the Public Debt," Journal of Political Economy, 87(5), 940-971. B RUMM , J. AND M. G RILL (2014): "Computing Equilibria in Dynamic Models with Occasionally Binding Constraints," Journal of Economic Dynamics and Control, 38(1), 142-160. C HARI , V.V. AND P.J. K EHOE (1990): "Sustainable Plans," Journal of Political Economy, 98(4), 784-802. C HARI , V.V. AND P.J. K EHOE (1993): "Sustainable Plans and Debt," Journal of Economic Theory, 61(2), 230-261. C HATTERJEE , S. AND B. E YIGUNGOR (2012): "Maturity, Indebtedness, and Default Risk," American Economic Review, 102(6), 2674-2699. 25 C OLE , H.L. AND T.J. K EHOE (2000): "Self-Fulfilling Debt Crises," Review of Economic Studies, 67(1), 91-116. C UADRA , G., J. S ANCHEZ , AND H. S APRIZA (2010): "Fiscal Policy and Default Risk in Emerging Markets," Review of Economic Dynamics, 13(2), 452-469. D’E RASMO , P. AND E.G. M ENDOZA (2013): "Optimal Domestic Sovereign Default," Mimeo, University of Maryland. D EBORTOLI , D. AND R. N UNES (2013): "Lack of Commitment and the Level of Debt," Journal of the European Economic Association, 11(5), 1053-1078. D IAZ -G IMENEZ , J., G. G IOVANNETTI , R. M ARIMON , AND P. T ELES (2008): "Nominal Debt as a Burden on Monetary Policy," Review of Economic Dynamics, 11(3), 493-514. D U , W. AND J. S CHREGER (2015): "Sovereign Risk, Currency Risk, and Corporate Balance Sheets," Mimeo, Federal Reserve Board. E ATON , J. AND M. G ERSOVITZ (1981): "Debt with Potential Repudiation: Theoretical and Empirical Analysis," Review of Economic Studies, 48(2), 289-309. G ARCIA , C. AND W. Z ANGWILL (1981): Pathway to Solutions, Fixed Points and Equilibria, Prentice Hall, Englewood Cliffs, NJ. H AMANN , F.A. (2004): "Sovereign Risk and Macroeconomic Fluctuations," Ph.D. Dissertation, North Carolina State University. H ATCHONDO , J.C., L. M ARTINEZ , AND H. S APRIZA (2010): "Quantitative Properties of Sovereign Default Models: Solution Methods Matter," Review of Economics Dynamics, 13(4), 919-933. H ATCHONDO , J.C., L. M ARTINEZ , AND C. S OSA -PADILLA (2015): "Debt Dilution and Sovereign Default Risk," Journal of Political Economy, Forthcoming. H UR , S., I. KONDO , AND F. P ERRI (2014): "Inflation, Debt, and Default," Mimeo, University of Pittsburgh. I LZETZKI , E. (2011): "Fiscal Policy and Debt Dynamics in Developing Countries," World Bank Policy Research Working Paper No. 5666. 26 K LEIN , P., P. K RUSELL , AND J. R IOS -RULL (2008): "Time-Consistent Public Policy," Review of Economic Studies, 75(3), 789-808. K RUSELL , P., F. M ARTIN , AND J. R IOS -RULL (2005): "Time-Consistent Debt," Mimeo, Princeton University. K RUSELL , P. AND S MITH , A.A. (2003): "Consumption–Savings Decisions with Quasi–Geometric Discounting," Econometrica, 71(1), 365-375. L IVSHITS , I., J. M AC G EE , AND M. T ERTILT (2007): "Consumer Bankruptcy: A Fresh Start," American Economic Review, 97(1), 402-418. L UCAS , R.E. AND N.L. S TOKEY (1983): "Optimal Fiscal and Monetary Policy in an Economy without Capital," Journal of Monetary Economics, 12(1), 55-93. M ARTIN , F.M. (2009): "A Positive Theory of Government Debt," Review of Economic Dynamics, 12(4), 608-631. M ARTIN , F.M. (2011): "On the Joint Determination of Fiscal and Monetary Policy," Journal of Monetary Economics, 58(2), 132-145. M ARTIN , F.M. (2013): "Government Policy in Monetary Economies," International Economic Review, 54(1), 185-217. M ARTIN , F.M. (2015): "Debt, Inflation and Central Bank Independence," European Economic Review, 79, 129-150. M ENDOZA , E.G. AND V.Z. Y UE (2012): "A General Equilibrium Model of Sovereign Default and Business Cycles," Quarterly Journal of Economics, 127(2), 889-946. NA , S., S. S CHMITT-G ROHÉ , M. U RIBE , AND V. Y UE (2015): "A Model of Twin Ds: Optimal Default and Devaluation," Mimeo, Columbia University. N IEMANN , S. (2011): "Dynamic Monetary-Fiscal Interactions and the Role of Monetary Conservatism," Journal of Monetary Economics, 58(3), 234-247. N IEMANN , S., P. P ICHLER , AND G. S ORGER (2013a): "Public Debt, Discretionary Policy, and Inflation Persistence," Journal of Economic Dynamics and Control, 37(6), 1097-1109. 27 N IEMANN , S., P. P ICHLER , AND G. S ORGER (2013b): "Central Bank Independence And The Monetary Instrument Problem," International Economic Review, 54(8), 1031-1055. N OCEDAL , J. AND S.J. W RIGHT (1999): Numerical Optimization, Springer Verlag, New York. PANIZZA , U., S TURZENEGGER , F. AND J. Z ETTELMEYER (2009): "The Economics and Law of Sovereign Debt and Default," Journal of Economic Literature, 47(3), 651-698. P OUZO , D. AND I. P RESNO (2014): "Optimal Taxation with Endogenous Default under Incomplete Markets," Mimeo, University of California, Berkeley. R EINHART, C.M. (2010): "This Time is Different Chartbook: Country Histories on Debt, Default, and Financial Crises," NBER Working Paper No. 15815. R EINHART, C.M. AND K.S. ROGOFF (2011): "The Forgotten History of Domestic Debt," Economic Journal, 121, 319-350. RÖTTGER , J. (2015): "Monetary Conservatism and Sovereign Default," Mimeo, University of Cologne. S OSA -PADILLA , C. (2013): "Sovereign Defaults and Banking Crises," Mimeo, McMaster University. S UNDER -P LASSMANN , L. (2014): "Inflation, Default, and the Denomination of Sovereign Debt," Mimeo, University of Copenhagen. S VENSSON , L. (1985): "Money and Asset Prices in a Cash-in-Advance Economy," Journal of Political Economy, 93(5), 919-944. T OMZ , M. AND M.L.J. W RIGHT (2007): "Do Countries Default in Bad Times?" Journal of the European Economic Association, 5(2-3), 352-360. 28 A A.1 Appendix Derivation of the Implementability Constraint I will only derive the implementability constraint for the repayment case. The constraint for the default case is derived similarly. First, take the household optimality conditions (1),(6)-(7) and rewrite them (in recursive notation) as un (n) 1 , u2 (c2 ) ψ(a, 0) u1 (1/p0 ) 1 , = β Ea0 |a u2 (c2 ) p0 u2 (c02 ) 1 − d 0 . = β Ea0 |a u2 (c2 ) p0 τ = 1+ 1+µ p (1 + µ) q p After using these expressions to eliminate the terms on the LHS of these equations in the government budget constraint g − τψ(a, 0)n + 1+B 1 + qB0 = (1 + µ) , p p one obtains un (n) 1 g− 1+ ψ(a, 0)n + 1/p + B/p u2 (c2 ) ψ(a, 0) u1 (1/p0 ) 1 u2 (c02 ) 1 − d 0 0 = β Ea0 |a + β Ea0 |a B, u2 (c2 ) p0 u2 (c2 ) p0 or un (n) n + 1/p + B/p g − ψ(a, 0)n − u2 (c2 ) u1 (1/p0 ) 1 u2 (c02 ) 1 − d 0 0 = β Ea0 |a + β Ea0 |a B. u2 (c2 ) p0 u2 (c2 ) p0 Now, eliminate ψ(a, 0)n via the resource constraint ψ(a, 0)n = 1/p + c2 + g, un (n) g − (1/p + c2 + g) − n + 1/p + B/p u2 (c2 ) u1 (1/p0 ) 1 u2 (c02 ) 1 − d 0 0 0 0 = β Ea |a + β Ea |a B. u2 (c2 ) p0 u2 (c2 ) p0 29 After multiplying both sides of the equation with u2 (c2 ) and using the policy functions to replace next period’s variables, one arrives at the implementability constraint −un (n)n − u2 (c2 ) (c2 − B/p) −1 1−D(B0 ,a0 ) r 0 0 0 0 u1 (P (B , a ) ) P r (B0 ,a0 ) r 0 0 1 − D (B , a ) = β Ea0 |a B0 . + β Ea0 |a u2 (C2 B , a ) r (B0 , a0 ) −1 D(B0 ,a0 ) P d 0 +u1 (P (a ) ) P d (a0 ) A.2 First-Order Conditions for the Policy Problem Conditional on repayment, the necessary first-order condition for an interior debt choice B0 is Z ∞ ∂ V r (B0 , a0 ) ∂ Rb 0 ∂ Rm b 0=ξ R + B + + β fa (a0 |a)da0 , ∂ B0 ∂ B0 ∂ B0 â(B0 ) with ξ denoting the multiplier on the implementability constraint (9).39 When combined with definition b u2 (c02 ) fa (a0 |a)da0 , 0 0 p â(B ) Z ∞ R =β and envelope condition ∂ V r (B, a) u2 (c2 ) = −ξ , ∂B p the first-order condition yields the generalized Euler equation Z ∞ β â(B0 ) u2 (c02 ) fa (a0 |a)da0 = ξ ξ −ξ p0 0 ∂ Rb 0 ∂ Rm B+ ∂ B0 ∂ B0 . The first-order conditions for the price index p, credit consumption c2 , and labor supply n are 0 = −p−2 [u1 (1/p) − ξ u2 (c2 )B − φ + ϑ u11 (1/p)] , (15) 0 = u2 (c2 ) + ξ [u22 (c2 ) (c2 − B/p) + u2 (c2 )] − φ − ϑ u22 (c2 ), (16) 0 = un (n) + ξ [unn (n)n + un (n)] + φ ψ(a, 0), (17) where ξ , φ and ϑ are the multipliers related to the constraints (9), (11) and (12). In addition to these 39 The derivation of the generalized Euler equation follows Martin (2009) and Arellano and Ramanarayanan (2012). 30 three conditions, the complementary slackness conditions ϑ ≥ 0, u1 (1/p) − u2 (c2 ) ≥ 0, ϑ × [u1 (1/p) − u2 (c2 )] = 0, (18) need to be satisfied as well. A.3 Numerical Solution The task of the numerical solution algorithm is to find the policy and value functions X r (B, a), X ∈ {B, C2 , N , P, V}, and X d (a), X ∈ {C2 , N , P, V}. Following Hatchondo et al. (2010), I approximate these functions on discrete grids for debt and productivity, and use cubic spline interpolation to allow for off-grid values of B and a. The solution algorithm involves the following steps: 1. Construct discrete grids for debt [B, B] and productivity [a, a]. r (B, a) and X d (a), X ∈{C , N , P, V}, 2. Choose initial values for the policy and value functions Xstart 2 start at all grid point combinations. j j 3. Set Xnext = Xstart , j ∈ {r, d} and fix an error tolerance ε. r (B, a), 4. For each discrete grid point combination (B, a) ∈ [B, B]×[a, a], find the optimal policies Xnew r (B, a). For each productivity X ∈ {B, C2 ,N , P}, and the associated value of repayment Vnew d (a), X ∈ {C , N , P}, and the value of devalue a ∈ [a, a], compute the optimal policies Xnew 2 d (a). fault Vnew r (B, a) − X r (B, a)| < ε and X d (a) − X d (a) < ε, X ∈ {C , N , P, V}, for all grid 5. If |Xnew 2 next new next j j point combinations, go to step 6, else set Xnext = Xnew , j ∈ {r, d} and repeat step 4. j 6. Use Xnew (·), j ∈ {r, d}, as approximations of the respective equilibrium objects in the infinite- horizon economy. For the debt grid, the individual points are distributed evenly. The bounds B and B are set such that they do not constrain the debt choice. Since the asymmetric default cost specification leads to a kink at a = ã in X d (a), X ∈ {C2 , N , P, V}, I partition the productivity grid for the default case as in Hatchondo et al. (2010) to account for this discontinuity. As is known in the literature (see e.g. Krusell and Smith, 2003, or Martin, 2009), there might be multiple Markov-perfect equilibria in models with infinitely-lived agents. In particular, there could 31 be equilibria with discontinuous policy functions which do not arise in the infinite-horizon limit of a finite-horizon model version. To avoid such equilibria, I follow Hatchondo et al. (2010) and solve for the infinite-horizon limit of a finite-horizon model version.40 In practice, this means that I compute the value and policy functions for the final period problem where no borrowing takes place and use j these objects as initial values Xstart , j ∈ {r, d}, for step 2. For a given state (B, a) ∈ [B, B] × [a, a], the objective function of the government is the sum of two parts, the period utility function u(1/p, c2 , n) and (in the repayment case) the continuation value r d (a) . The optimal policies for step 4 β Ea0 |a [Vnext (B0 , a0 )], with Vnext (B, a) = max Vnext (B, a), Vnext are then computed as follows. I use a sub-routine that calculates the optimal static policies c2 , n, and p for given debt and productivity values (B, a) ∈ [B, B] × [a, a] and an arbitrary, i.e. possibly off-grid, borrowing value B̂0 . More specifically, I use a non-linear equation solver to find the variables that satisfy the static optimality conditions (9),(11),(15)-(18).41 Since these conditions involve the complementary slackness conditions (18), I follow Brumm and Grill (2014) and use the trick by Garcia and Zangwill (1981) to transform the set of optimality conditions into a system of equations.42 Using the static policy sub-routine, (c2 , n, p) and thus period utility u(1/p, c2 , n) can be expressed as functions of (B, a, B̂0 ).43 As a result, given (B, a) ∈ [B, B] × [a, a], the government objective can be expressed as a function of B̂0 as well: u(1/p, c2 , n) + β Ea0 |a Vnext (B̂0 , a0 ) . r (B, a) For each discrete grid point combination (B, a) ∈ [B, B] × [a, a], the optimal debt policy Bnew then is computed via a global non-linear optimizer, calling the static policy routine to calculate the objective function for each candidate debt value B̂0 . More specifically, for each (B, a), I first perform a grid search over a pre-defined grid for B̂0 and then use the solution as an initial guess for the NelderMead algorithm.44 The optimal policies X r (B, a), X ∈ {C2 ,N , P} then are computed by using the r (B, a). The algorithm iterates on the polstatic policy routine for the optimal borrowing value Bnew icy and value functions until the maximum absolute difference between value and policy functions 40 Martin (2009) also solves for the infinite-horizon limit. As pointed out by him, using a Svensson (1985)-type beginningof-period cash-in-advance constraint in a finite-horizon model requires a terminal money value for a monetary equilibrium to exist. Otherwise, households will not be willing to invest in money in the final period and by backward induction not in any of the previous periods. The impact of the final-period value of money vanishes over time and does not affect the final results. 41 In the default case, condition (9) is replaced by condition (10) 42 Alternatively, one can also use a non-linear constrained optimizer to compute the optimal static policies for each combination (B, a, B̂0 ). Using a sequential quadratic programming algorithm (see e.g. Nocedal and Wright, 1999 for details), I found this approach to be both slower and less accurate. 43 The routine is also used to obtain the optimal policies in the final period, where B̂0 = 0 holds. 44 Instead of the Nelder-Mead algorithm, I also solved for the optimal debt policy using Golden section search and a mesh adaptive direct search algorithm (as implemented by the optimizing routine NOMAD provided by the OPTI Toolbox), which did not affect the results. 32 obtained in two subsequent iterations is below ε = 10−5 for each combination (B, a) ∈ [B, B]×[a, a].45 To evaluate value and policy functions at debt and productivity states that are off-grid, cubic spline interpolation is used.46 To approximate expected values in an accurate way, one needs to account for the default threshold. This can be seen by looking at the expected option value of default: Z Ea0 |a Vnext (B0 , a0 ) = 0 â(B0 ) d Vnext (a0 ) fa (a0 |a)da0 + Z ∞ â(B0 ) r Vnext (B0 , a0 ) fa (a0 |a)da0 . Gauss-Legendre quadrature nodes and weights are used to approximate the integrals above. The r (B, â(B))−V d (â(B)) = 0 and is computed via bisection method. default threshold â(B) satisfies Vnext next 45 Using a tighter convergence criterion did not affect the results. 46 Hatchondo et al. (2010) show that allowing for a continuous state space is crucial for accurate solutions of models with equilibrium default. 33